VLSI Design Project Audio Steganography - PowerPoint PPT Presentation

1 / 13
About This Presentation
Title:

VLSI Design Project Audio Steganography

Description:

Default power analysis is done by probabilistic flipping of ... CRP. 191822. DES3. 206606. CODEC (V0) Area (um^2) 0.165. 2.832. 1.186. 1.647. SPI (V1) 0.250 ... – PowerPoint PPT presentation

Number of Views:915
Avg rating:3.0/5.0
Slides: 14
Provided by: bendo9
Category:

less

Transcript and Presenter's Notes

Title: VLSI Design Project Audio Steganography


1
VLSI Design ProjectAudio Steganography
  • Implementation Review Slideshow
  • Ben Doherty Chris Campetti
  • Jeremy Gummeson Phil Murray

2
Statistics
  • PRNG Auto-Correlation
  • 10,000 samples
  • Maximum over a span of 1000 with a lag of 1
    0.0363
  • This indicates very little correlation
  • Power consumption for entire design
  • Probabilistic values given by Power Compiler
  • Dynamic 384.228mW
  • Leakage 8.730uW
  • Power is most heavily consumed in register file
    (157mW/6.8uW) and 3DES core (136mW/0.5uW)

3
Initial Power Analysis
  • Synopsys Power Compiler
  • Default power analysis is done by probabilistic
    flipping of nodes throughout design
  • Good for average case scenarios, but has little
    to do with power consumption during usage
  • Still trying to flesh out input vector methods
  • SAIF files Backward/Forward SAIF
  • Verilog-XL may be able to generate these

4
Design Iterations SPI
  • First design (V0) had a single FSM, which was
    clocked on both SPI_CLK edges
  • Would verify under Verilog-XL, but not synthesize
    properly (lack of dual-edge registers)
  • Second design split FSM into two separate FSMs
  • One operated Regfile control signals, the other
    shifted bits in/out
  • Significant improvement of power/area

5
Design Iterations Register File
  • V0 wasted a lot of area on multiple large
    multiplexors
  • always _at_(posedge write_latch) begin
  • if(REG_wr0 1'b1)
  • reg_wrdata lt REG_write0
  • else if(REG_wr1 1'b1)
  • reg_wrdata lt REG_write1
  • end
  • V1 had code to EXPLICITLY infer specific
    multiplexors for input data registers
  • wire 10 wrdata_MUX_sel
  • assign wrdata_MUX_sel REG_wr0,REG_wr1
  • wire FILEWIDTH-10 wrdata_mux
  • assign wrdata_mux (wrdata_MUX_sel 2'b00 ?
    0
  • wrdata_MUX_sel 2'b01 ? REG_write0
  • wrdata_MUX_sel 2'b10 ? REG_write1 0)
  • // The write data latch register
  • always _at_(posedge write_latch)

6
Design Iterations - PRNG
  • V0 Multiple Driver issue on load register
  • Again, would simulate properly in Verilog-XL, but
    would cause synthesis errors during compile
    (Assumed a wired-AND)
  • V1 Inferred shift register with asynchronous
    load
  • always _at_(posedge read_in or posedge clk) begin
  • if(read_in 1'b1)
  • shift lt key_in

7
Area/Power Contributions
POWER
AREA
8
Cell References
9
Who Did What?
  • Ben
  • REGFILE and SPI interface, PRNG V1
  • Verification test benches
  • Design analysis power, area, cell references
  • Chris
  • CODEC
  • Initial PRNG
  • Test bench and Analysis
  • Preliminary design analysis
  • Jeremy
  • Mostly CODEC module development
  • Verification of CODEC
  • Phil
  • Layout, schematics
  • Some design analysis

10
Design Flow
11
Modular Floorplan
12
What we have so far
  • Complete design, including
  • Pseudo-dual port register file memory
  • SPI Interface
  • Steganography CODEC
  • PRN generator
  • 3DES module from OpenCores
  • Design has been laid out and verified

13
What we plan on doing
  • FPGA implementation on DE2 board
  • Modular IC layout
  • More thorough power analysis
  • More thorough verification
  • RTL
  • CODEC bugs/improvements (V2)
Write a Comment
User Comments (0)
About PowerShow.com